Choosing camera resolution for dimensional inspection

A higher megapixel count does not, by itself, make a reliable dimensional inspection. You need enough pixels across the permitted dimensional variation, stable edges, controlled perspective and an exposure short enough to freeze conveyor motion. This guide shows how to specify those requirements for coated bars before choosing hardware. The worked example uses hypothetical dimensions and acceptance limits; replace them with your drawing, production variation and validated measurement requirements.

Define the dimensions before choosing the camera

Start with the inspection decision, not the existing camera resolution. Record the nominal length and width, permitted deviations, product orientation, conveyor speed and full placement envelope. Also define whether the measurement concerns the top surface, outer silhouette or another specified datum.

For the worked example, assume:

• Nominal bar dimensions: 170 mm long, 35 mm wide and 35 mm high.
• Length and width acceptance limits: nominal dimension ±0.5 mm.
• Required image coverage: 240 mm horizontally and 160 mm vertically.
• Conveyor speed during exposure: 800 mm/s.
• Initial sampling target: 0.1 mm per pixel at the measurement plane.

These are application assumptions, not specifications of any catalog product. The field of view must include the complete bar throughout its allowed translation and rotation, with space for edge searches.

Do not substitute blob area for separate dimensional measurements. A rectangle measuring 170 × 35 mm has an area of 5,950 mm². Another measuring 175 × 34 mm has the same area, despite having different length and width.

Agree how to treat coating drips, rounded ends and chipped corners before establishing rejection limits.

Translate tolerance into object-space sampling

Object-space sampling describes how much of the product each image pixel represents. It is separate from accuracy: more pixels cannot remove calibration bias, perspective errors or unstable lighting.

Here, p is measured in millimeters per pixel, and N is the required number of pixels along an image axis.

As an initial engineering target, sample the smallest relevant dimensional deviation with several pixels rather than one. For this example, a 0.5 mm deviation spanning five pixels gives:

Five pixels is a starting assumption, not a universal measurement standard. Required sampling depends on edge contrast, the measurement algorithm and the uncertainty budget. A repeatable subpixel fit can improve edge localization, but its performance needs evidence from representative images.

A length measurement uses two edges. If each edge can shift by 0.5 pixel, the worst-case opposing shifts produce a one-pixel length change. At 0.1 mm/pixel, that contribution is 0.1 mm before other errors are considered.

Work through the field-of-view requirement

Apply the sampling target separately to both image axes:

The result is approximately 3.84 megapixels, but that total alone is insufficient. The camera must provide at least the required pixel count along each axis in the acquisition mode you will actually use. A different sensor aspect ratio can provide enough total pixels while missing one axis requirement.

Use the camera finder to shortlist candidates, then check their exact active image dimensions. The resolution calculator provides a starting point for the sampling calculation.

At the assumed sampling, the nominal bar spans 1,700 pixels in length and 350 pixels in width. Those figures help assess whether the edge-search windows are sensible and whether the proposed image processing load is practical.

Check the placement envelope as well. A rotated rectangular bar occupies more vertical image space than its nominal width suggests. The calibration target coverage guide explains why useful coverage and calibration margins must be planned together.

Choose the camera by shutter and acquisition mode

For continuously moving products, a global shutter is generally the straightforward starting point. It exposes the image without the row-dependent timing that can distort moving objects with a rolling shutter. It does not eliminate motion blur; exposure duration still matters.

Browse area scan cameras using the calculated axis requirements, then consider shutter type, interface, trigger support and sustainable acquisition rate. Confirm every required acquisition feature in the selected product documentation.

The MV-CH120-11UC camera is a global-shutter, color candidate for evaluation. Its catalog resolution makes it worth investigating, but exact image dimensions and acquisition capabilities still need checking against the application.

A color camera can measure dimensions, but demosaicing and color-dependent contrast may affect edge localization. Validate the actual image format and processing path rather than assuming every output pixel provides an independent intensity measurement. A monochrome option may be preferable when color contributes nothing to the inspection.

Check the complete acquisition chain: camera, host interface, cable, driver and inspection software. A camera upgrade is useful only if the installed system can acquire and process its images consistently. The camera upgrade compatibility guide covers the checks that resolution changes can introduce.

Choose optics and control height sensitivity

The lens must cover the sensor, provide the required field of view and resolve the edges across the measurement region. Matching the mechanical mount is necessary, but it does not establish optical suitability.

Use machine vision lenses and the lens finder to compare coverage and focal-length options. The focal-length selection guide explains the relationship between sensor dimensions, working distance and image coverage.

The LS-FFL-0567 lens is a candidate for compatibility review. Its mount and listed sensor format make it relevant to the camera shortlist, but the required field of view must still be checked at the proposed mounting distance.

Ordinary perspective lenses change apparent scale when the measurement surface changes height. Calibrating on the belt does not automatically establish the correct scale on the top of a raised product.

For an approximate pinhole model, let D be the camera-center distance to the reference plane and h the product height above it. Small height changes produce approximately:

For an illustrative reference distance of 500 mm, nominal height of 35 mm and height variation of 2 mm:

That exceeds the example dimensional tolerance. Adding pixels would not solve this geometric problem. Stabilize product height, calibrate at the actual measurement plane, increase viewing distance where practical, or evaluate object-space telecentric optics. Products with substantial height variation may require height measurement and geometric compensation instead.

Form an edge that remains stable

For an opaque object with suitable optical access, backlighting often produces a useful silhouette. However, an opaque conveyor belt blocks light from below. A backlight requires a suitable transparent support or an approved transfer arrangement, not simply a brighter light.

Compare backlights through the lighting finder. Check illuminated coverage, mounting space, support shadows, product position and compatibility with the acquisition arrangement.

The LS-BKL-0869 backlight is a candidate to review for coverage and installation. Confirm its usable emitting area, electrical requirements and supported operating modes before selection.

The diffuse versus collimated backlighting comparison helps distinguish arrangements that produce different edge profiles. Evaluate them using measured edge position, not just which image looks sharper.

When backlighting is unavailable, reflected illumination can separate a dark product from a contrasting belt. Glossy coating may create highlights that split the thresholded region or move the apparent boundary. Compare edge profiles across representative coating conditions and belt contamination levels.

Keep exposure, gain and lighting settings fixed during capability testing unless adaptive behavior is part of the validated design. Automatic brightness adjustments can change threshold crossings and make apparent dimensions drift between otherwise similar products.

Set exposure from the blur budget

Calculate exposure from allowable movement during exposure, not from frame rate alone. A camera can deliver images frequently while each individual image remains blurred.

For the example speed, sampling and a provisional blur allowance of 0.5 pixel:

An exposure near 60 microseconds is therefore an initial test setting, not a guaranteed optimum. Verify that edge contrast remains adequate without excessive gain. Actual blur tolerance depends on the edge estimator and required measurement uncertainty.

If continuous illumination cannot support the necessary exposure, investigate a documented pulsed-light arrangement with compatible timing and electrical limits. Do not infer pulse capability from a light's general description.

The conveyor camera selection guide provides additional context for balancing sampling, motion and acquisition rate. Confirm that the inspection produces a valid decision for every product at maximum line speed.

Demonstrate capability before setting rejection limits

Calibrate with a suitable dimensional reference at the intended measurement plane. Check scale and residual error across the field rather than validating only the center. The image-marker measurement calibration guide explains how pixel measurements become physical dimensions.

Build a validation set containing known dimensions, representative surface finishes, different orientations and expected height variation. Include products near both acceptance boundaries, not just obvious rejects.

Separate these checks:

• Repeatability: repeated measurements of an unchanged, stationary reference.
• Position sensitivity: the same reference measured across the usable field.
• Dynamic performance: comparable measurements while the reference moves.
• Reproducibility: results after normal restarts, cleaning and approved setup adjustments.
• Bias: disagreement between vision measurements and a suitable reference measurement.

Suppose validation establishes an expanded measurement uncertainty of 0.2 mm under the documented operating conditions. With length specifications of 169.5–170.5 mm, a conservative full-uncertainty guard band would give an automatic acceptance interval of 169.7–170.3 mm.

Measurements outside that guarded interval need the agreed disposition: rejection, diversion or further inspection. The guard band reduces false-accept risk but can increase false rejects. It is an example decision rule, not a substitute for the quality plan.

Record the uncertainty basis and decision rule explicitly. Camera sampling is only one input; reference uncertainty, calibration, edge fitting, height variation and lighting stability can dominate the result.

Check the specification before ordering

The usual answer of buying more pixels is wrong when the limiting problem is unstable geometry, inadequate contrast or undefined measurement rules. Resolve those issues before committing to a larger camera.

Use this final checklist:

• Confirm separate length and width limits, including treatment of drips and broken edges.
• Document the full placement envelope, rotation range and actual measurement height.
• Calculate required pixel counts along both axes, not just total megapixels.
• Verify sensor coverage, lens mount, working distance and corner image quality.
• Test the lighting arrangement with representative products and conveyor supports.
• Calculate an exposure limit and check dynamic measurements at maximum speed.
• Confirm interface compatibility, trigger behavior and sustained processing capacity.
• Approve the calibration procedure, uncertainty budget and rejection decision rule.
• Validate rejection mechanics and machine safety separately from image measurement.

Begin the camera configuration process once the sampling and acquisition requirements are documented. Continue with the lens configuration step and lighting configuration step to review the complete image-forming arrangement. For a selection review, send the dimensional drawing, field of view, speed, height range and representative images through application support.